Framework for developing quantitative agent based models based on qualitative expert knowledge: an organised crime use-case

• In Event: Methodology and law enforcement

Abstract

Objectives Developing Agent-Based Models (ABMs) of organized crime network dynamics is a promising approach to support the design of predictions and interventional strategies for law enforcement. The problem is that too little quantitative data is available to effectively identify a unique ABM model, which is also true in many other psychosocial contexts.
Methods Here, we propose a modelling framework that also incorporates qualitative data, such as police reports, literature, and expert interviews, to systematically develop, train, and validate computational ABMs in data-poor contexts. In our framework, first a conceptual model is designed in close collaboration with domain experts. Then, the conceptual model is then refined and translated into a quantitative model through thematic content analysis based on both the quantitative and qualitative data analysis. Finally, the quantitative model is calibrated and validated by supplementing the quantitative data with training and validation statements, which are distilled from the qualitative data. This is the most innovative part of our framework. Lastly, scenario testing, sensitivity analysis and uncertainty quantification complete the modelling cycle. We illustrate this framework through an exemplary case study of the criminal cocaine network in the Netherlands.
Results The resulting ABM, called Criminal Cocaine Replacement Model (CCRM), captures dynamics of kingpin removal and replacement.
Conclusion Our scenario testing confirms that the proposed framework enables the development of accurate ABMs in the case of little qualitative data.

Authors

• Frederike Oetker, University of Amsterdam

• Vittorio Nespeca, University of Amsterdam

• Thijs Vis, Police Amsterdam

• Paul Duijn, Institute for Advanced Science, University of Amsterdam

• Peter Sloot, University of Amsterdam

• Rick Quax, University of Amsterdam